Lists Home |
Date Index |
Martin Klang wrote:
> let me see if i got this right -> if i use/write/promote a tool based on a limited
> datamodel, i'm threating the future of XML??
No. If, however, you promote an *XML* based on a limited datamodel you are threatening
not just the future but the present of XML (and for that matter the backward-,
legacy-compatible past). I am not speaking of a use, nor an application, nor a profile
of XML here, but of the concept of XML itself. The promise of XML is interoperability.
That is a subtle thing, which occurs in different ways and in different levels at once.
Within the document there is interaction among the elements, and further interaction
among all of the identifiable markup artifacts. Interactions take place in a particular
instantiated instance, and those particular interactions, like that instance, may well
be unique. That instantiation is built on the output of a parse--a parse of a specific
XML document in particular circumstances on a specific occasion by a specific parser,
and then manipulated by a particular processor to produce a particular data
instantiation from the parse. There is no inherent nor persistent connection between the
XML document parsed and the data instantiation ultimately built on a given occasion upon
the output of that parse. This means that the interrelationship and the interactions of
the elements and other markup artifacts of an XML document will vary with each
instantiation of data from that document. Likewise the relationships between documents
or between versions of a given document will vary as differing datastructures are
instantiated from those documents on various occasions out of independent parses.
Within all of this difference, where does the basis for interoperability lie? In the XML
syntax of the documents themselves. It is precisely because XML documents can be used by
different processes as the basis for entirely different function that such a document
can be the nexus of interoperability between dissimilar, autonomous, and perhaps
mutually anonymous processes. A datamodel, however, is nothing more than the abstraction
of the particular data instantiation built on a particular occasion on the output of a
particular parse of a given XML document. The document is the only thing in that chain
of process with more than ephemeral existence. Any particular data instantiation is no
more than an accident of the specific circumstances of a given parse, and then of the
particular manipulation of the output of that parse. A datamodel which is the schematic
of that data instance is no more than the abstraction of that accident.